Skip to main content

How to use product recommendation chat so shoppers pick the right item

Poor fit is the top reason shoppers send items back, and product recommendation chat that asks a few questions before suggesting anything catches the mismatch before checkout.

Key takeaways

  • Poor fit is the top return reason, named by 61% of shoppers in Rithum's 2025 global survey.
  • Ask how the shopper will use the item before you suggest anything, then recommend one pick with a reason.
  • Saved answers cover repeat sizing questions instantly but cannot weigh several details at once.
  • An AI agent answers at any hour, but only as well as your product data allows.
  • Hand fit questions the data can't answer to a person, and say who is replying.

Good product recommendation chat finds a shopper the right size or model with a few questions, not a pitch. Poor fit is the top return reason, named by 61% of shoppers in Rithum’s 2025 global survey. Here’s how to answer before the wrong item ships.

Below you’ll find why shoppers ask, what to ask them back, and three ways to answer at scale: doing it yourself, saved answers and an AI agent that reads your catalog. Each comes with what it costs you in time and where it falls short.

Why do shoppers ask product questions before they buy?

Shoppers ask because the product page left a gap, usually about fit, size or how two similar items differ. In Syndigo’s 2024 State of Product Content survey of adults in the US, France, Germany and the UK, 50% had walked away from a purchase at some point over the prior half year when the details they wanted weren’t there. In the same survey, 83% said they were likely to leave a site with thin product content.

A woman at a desk takes a phone photo of a black sneaker, with a laptop and shipping boxes beside her
Photo: Kampus Production (Pexels License)

A visitor who leaves is a lost sale. A visitor who guesses wrong is a return. Poor fit is the most cited return reason among the more than 6,000 consumers in Rithum’s 2025 global survey, and 68% of them had sent back an item of apparel or footwear within the last year. Syndigo found that 35% had returned a product because it didn’t match what the product content led them to expect.

Some shoppers settle the doubt by ordering two sizes and sending one back. Rithum calls this bracketing, buying extra on purpose to return part of the order, and 36% of shoppers admitted to it. Among shoppers under 35, it was 50%. A clear answer in chat gives them a reason to order one size.

Most shoppers never write in on their own, so the chat has to make asking easy. During the 2025 Cyber 5 holiday sales, just 5.8% of the 1,022 shoppers in a Digital Commerce 360 and Bizrate Insights survey reached out to a store’s support team. A chat window on the product page, with the common questions ready to tap, reaches the ones who would otherwise guess.

How do you guide a shopper to the right item without pushing?

Ask how they’ll use the item before you suggest anything, then recommend one product and say why it fits. Questions show the shopper you’re solving their problem. A pitch shows them you’re solving yours.

A smiling seller writes in a notebook beside a laptop, with cardboard boxes and a clothing rack around her
Photo: Kampus Production (Pexels License)

Pick one or two of these questions and stop once you know enough to recommend:

  • What will you mainly use it for?
  • What do you have now, and what would you change about it?
  • Which size or model fits you in another brand?
  • Is there a price you’d like to stay under?

Sizing deserves the most care because so many shoppers come looking for it. In a 2021 survey of 7,528 US shoppers by PowerReviews, 58% of those who read the questions and answers on a product page were looking for sizing information. Ask for a measurement or a size they wear in another brand, then check it against your size chart before you reply.

When you recommend, give one main pick and at most one alternative, each with a reason in the shopper’s own terms. A shopper who said they run cold should hear that the lined version suits them better, and why. Then stop. If they want more options, they’ll ask.

Say so when a product isn’t right for them. Telling a shopper that a jacket runs small, or that the cheaper model does everything they need, costs you an upsell. It also heads off a return and the shipping both ways.

Point to what other buyers said. Half of consumers lean hard on reviews before ordering apparel or footwear, according to Rithum, so a note that reviewers find a shoe runs large carries weight your own word may not.

Tell shoppers who’s answering. In the 2021 PowerReviews survey, 82% of US shoppers said they notice the line showing who replied to a product question. A first name, or a plain note that a bot is replying, sets the right level of trust.

Keep each reply short enough to read on a phone. Salesforce data reported by MarTech shows phones and tablets carried 74% of holiday orders worldwide in 2025, compared with 62% a year earlier. Two or three sentences and a link to the product page read better on a small screen than a block of text.

Answer return questions plainly, because a vague policy can cost you the customer. In Rithum’s survey, 47% of consumers said they’d quit buying from a store over a return policy that disappointed them. Our guide to handling a return request in chat covers the replies for when an item does come back.

See how LenoChat handles this for small teams.

What are your options for answering product questions over chat?

You have three: answer every question yourself, write saved answers for the questions that repeat, or let an AI agent answer from your product catalog. You can combine them, so treat each as a piece you add rather than a single choice.

Answer every question yourself

Answering by hand gives the best fit advice, because you can weigh several details at once: the shopper’s height, the brand they usually wear and what they’ll use the item for. You know which jacket runs small and which leather stretches. A script or a bot only knows that if you’ve written it down.

The cost is your hours. A person answers only while someone watches the inbox. In a 2021 PowerReviews survey, 56% of US shoppers wanted their product question answered inside 24 hours. A shopper typing into a chat window is still on the product page, deciding, so the reply has to land while they’re there.

Answering by hand goes faster when you can see what the shopper is looking at. A chat tool connected to your store can put the shopper’s cart and the product details beside the conversation, so you can suggest a size with both in front of you. The Shopify live chat integration page explains what that connection shows.

Write saved answers for the questions that repeat

Saved answers, replies you write once and send with a click, handle the questions that come up every week: the size chart, whether a style runs small, how a fabric feels, which model a part fits. Set the most common ones as conversation starters, buttons in the chat window that a shopper taps to see the answer right away, and they keep working when nobody is online.

A man in a workshop apron sits at a laptop and writes notes on a pad
Photo: Antoni Shkraba (Pexels License)

Start with sizing, since poor fit is the reason behind the most returns and it has a fixed answer you can write once. Give the measurement, how the item fits next to a standard size and what to do if the shopper is between sizes.

Saved answers have one limit: they can’t combine details. A shopper who asks whether the medium works for someone tall with long arms needs judgment, and a fixed reply sends them the size chart again. End each sizing answer with a line saying a person can help if they’re between sizes, and route those chats to whoever knows the products.

Keep them current. A saved answer that names a discontinued color or an old size chart creates the very mismatch it was written to prevent. Small repeat questions are often the ones closest to a sale, as our piece on turning simple shop questions into sales explains.

Let an AI agent answer from your catalog

An AI agent, a chatbot that writes its own reply from your product data and help pages, answers at any hour and handles questions you never wrote an answer for. Connected to your catalog, it can tell a shopper which models come in their size and which are in stock.

Shoppers already ask AI for this kind of help. In Adobe’s March 2025 survey of more than 5,000 US consumers, 39% had used generative AI, tools that write an answer to a question, for online shopping, and 47% of those users had asked it for product recommendations. Most of them, 87%, said they’d reach for it sooner on a bigger or trickier buy.

Retailers are adding AI agents for the busiest weeks. Service conversations handled by AI rose 66% in December 2025, according to Salesforce data reported by MarTech. Over the same holiday season, sales grew 32% faster at retailers running a branded AI agent than at retailers without one. That shows the two go together, not that the agent caused the growth.

The agent is only as good as your product pages. If a description doesn’t say a boot runs narrow, the agent can’t say it either. Before you switch one on, add measurements, fit notes and compatibility details to your descriptions, then test it with the questions shoppers actually send. Set it to pass any fit question it can’t answer confidently to a person, with the conversation so far attached.

A middle path is to choose the products yourself and let the agent show them. In LenoChat, product recommendations in chat let you build up to 5 curated sets of 3 to 10 products, each tied to up to 10 search terms, and the AI agent displays a set as pictures, product names and store links inside the conversation. It needs a connected Shopify catalog, a set takes under 2 minutes to build, and every automated reply references the source it drew on.

A chat inbox listing shopper conversations tagged Product Request and New Arrival, with a reply suggesting an item and its price

If you’re still deciding between a bot and a person, our comparison of AI chatbots and live chat for small business sales weighs both.

How do the three options compare?

A person gives the best fit advice but only during your hours, saved answers reply at once but only to what you wrote, and an AI agent covers every hour for anything your product data explains. Reply speed matters for all three: in the 2021 PowerReviews survey, 21% of US shoppers wanted a reply to a product question in four hours or less.

OptionHours coveredFirst replyQuestions it can answerWork before launchOngoing work
Answer every question yourselfStaffed hours onlyMinutes to hours, depending on who is freeAny, including fit across brandsNoneEvery chat, by hand
Saved answersAll hours, when set as tap buttonsImmediateOnly the ones you wroteOne answer per repeat questionRewrite when products or size charts change
AI agent on your catalogAll hoursImmediateAnything your product data and help pages explainConnect the catalog, fill gaps in product pages, testReview unanswered questions and fix the data

Saved answers take less upkeep than an AI agent when you sell a small range that rarely changes, and an AI agent answers more questions than saved answers once your catalog has more products than you can write answers for.

Answering at all is what pays. PowerReviews reported in 2021 that site visitors who used product questions and answers converted, meaning they bought, 157.1% more often than visitors who didn’t, based on its own analysis. That measured questions on the product page, not chat, and shoppers who read them may have been closer to buying already, but it points the same way.

Which option should your store start with?

Start with saved answers for your sizing and fit questions, answer everything else yourself, and add an AI agent once questions start arriving while you’re closed.

That order follows the cost of getting it wrong. Saved answers are quick to write and fix the question behind the most returns. Answering the rest by hand shows you which questions repeat, which tells you what to add to your product pages. An AI agent comes last because it answers from those pages, so it’s only as good as the work you did first. If most of your chats already come in after hours, move it up, and read our piece on selling through chat when you’re busy or offline for the setup around it.

Answer every question yourself. Best for stores with a small catalog and fit questions that need judgment. Not for stores that get most of their chats after closing.

Saved answers. Best for stores where the same sizing and material questions repeat on every product. Not for questions that combine several details about the shopper.

AI agent on your catalog. Best for stores with a large catalog or most questions arriving after hours. Not for stores whose product pages are missing measurements and fit notes.

Whichever you pick, look once a week at the product questions that went unanswered or needed a person, and add the missing detail to the product page. Each fact you put on the page is one less question in chat and one less wrong size in the mail.

Ready to answer every customer from one place? Start free with LenoChat.

Frequently asked questions

Can a chatbot tell a shopper which size to buy?

Only when your product data includes measurements and fit notes, such as whether a style runs small. Without them, the bot can repeat the size chart but can't judge fit, so set it to hand those questions to a person.

How many products should you suggest in one chat reply?

One main pick and at most one alternative. Each suggestion should come with a reason tied to what the shopper told you. A longer list hands the decision back to them and slows the sale.

What should a product page include before you turn on an AI chat agent?

Measurements, fit notes, materials, what's in the box and what the item works with. An AI agent answers from that text, so a gap on the page becomes a gap in its answers.

Should shoppers know whether a person or a bot is answering?

Yes. In a 2021 PowerReviews survey, 82% of US shoppers said they notice who answered a product question. A first name, or a clear note that a bot is replying, sets honest expectations.

Sources

  1. Rithum
  2. Digital Commerce 360
  3. MarTech
  4. Adobe
  5. Syndigo
  6. PowerReviews

Image credits

  • Photo: Vitaly Gariev (pexels)
  • Photo: Kampus Production (pexels)
  • Photo: Kampus Production (pexels)
  • Photo: Antoni Shkraba (pexels)

Related Articles

Store owner filling in order paperwork at a desk stacked with packed shipping boxesMarketing

How to handle an order cancellation request chat without losing the customer

Most order changes and cancellations take one reply, once you know whether the parcel has shipped. Wrong addresses are costly: in Sendcloud’s 2025 survey of UK shoppers, 64.3% said a failed delivery would stop them reordering. Here’s how to run each order cancellation request chat. These chats come in three shapes: fix the address, swap […]

Two customer support agents talking while working at the same deskMarketing

How to run agency live chat for several clients

Table of contents How do you keep each client separate in one inbox? Who should see each client’s conversations? How do agents write in each client’s brand voice? How do you keep service consistent across clients? Which inbox setup should your agency choose? Agency live chat can cover several client brands without mixing conversations or […]

A support agent wearing a headset types a reply on a laptop at an office deskMarketing

Apology message templates for wrong items, late orders, double charges and missed replies

A good apology message template lets you say sorry fast without sounding like a form letter. Only 40% of US complainants came away fully satisfied with how their complaint was handled, according to a 2025 survey by Customer Care Measurement & Consulting. Below are ready-to-paste apologies for four common mistakes. What should every apology message […]